I find that VERY doubtful. Take the AT review of Metro 2033 at 1200p for instance. 54 fps vs. 44.5 fps. And then look at system power draw: 391 vs. 425 watts. So you have 21.3% more performance and 8% less power draw. Overall perf/watt goes up ~30%
even if you look at the entire system.
http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/27
http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/17
And to get a 72% increase in perf/watt over the GTX 580 when the performance is only ~20% faster than the 7970 would imply a laughable power draw.
Edit to add: I don't mean that I find your logic doubtful, just that even using system power draws and assuming roughly comparable test machines, it's pretty clear that 7970 is more than 17% more efficient than 580 even if you look at system total power draw rather than the card itself. Something doesn't add up. The chart can't be right... maybe the 680 isn't as fast as the charts say, or it eats up way more power, or something, but 72% more efficient than the 580 even at the system level (which would imply what, almost DOUBLE the efficiency of the 580 when comparing card vs card?) is outrageous and would mean a HUGE coup by Nvidia.
Given the obvious error when comparing the 7970 to the 580, I think the simpler explanation is that one or more of these slides is or are wrong.